With the currently ongoing and necessary process to define standards for multi-agent systems, their potential for adaptation to specific underlying system requirements becomes increasingly challenging. Especially in those applications, where multi-agent systems are coupled to hard- and software systems, which have specific operational characteristics, such as specific response times or quality-of-service requirements. As a result, it is proposed to separate the high-level decision making based on standardized multi-agent systems and the low-level system control into two layers. In this paper, based on a publisher/subscriber software-bus system, we propose a framework which allows dynamic allocation and linking of agents to underlying low-level hardware control. The concept of the framework and its architecture is introduced and an application example is outlined.